1/*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (c) 1982, 1986, 1993, 1994, 1995
5 * The Regents of the University of California. All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the University nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * SUCH DAMAGE.
30 */
31
32#ifndef _NETINET_TCP_VAR_H_
33#define _NETINET_TCP_VAR_H_
34
35#include <netinet/tcp.h>
36#include <netinet/tcp_fsm.h>
37
38#ifdef _KERNEL
39#include <net/vnet.h>
40#include <sys/mbuf.h>
41#include <sys/ktls.h>
42#endif
43
44#define TCP_END_BYTE_INFO 8 /* Bytes that makeup the "end information array" */
45/* Types of ending byte info */
46#define TCP_EI_EMPTY_SLOT 0
47#define TCP_EI_STATUS_CLIENT_FIN 0x1
48#define TCP_EI_STATUS_CLIENT_RST 0x2
49#define TCP_EI_STATUS_SERVER_FIN 0x3
50#define TCP_EI_STATUS_SERVER_RST 0x4
51#define TCP_EI_STATUS_RETRAN 0x5
52#define TCP_EI_STATUS_PROGRESS 0x6
53#define TCP_EI_STATUS_PERSIST_MAX 0x7
54#define TCP_EI_STATUS_KEEP_MAX 0x8
55#define TCP_EI_STATUS_DATA_A_CLOSE 0x9
56#define TCP_EI_STATUS_RST_IN_FRONT 0xa
57#define TCP_EI_STATUS_2MSL 0xb
58#define TCP_EI_STATUS_MAX_VALUE 0xb
59
60#define TCP_TRK_REQ_LOG_NEW 0x01
61#define TCP_TRK_REQ_LOG_COMPLETE 0x02
62#define TCP_TRK_REQ_LOG_FREED 0x03
63#define TCP_TRK_REQ_LOG_ALLOCFAIL 0x04
64#define TCP_TRK_REQ_LOG_MOREYET 0x05
65#define TCP_TRK_REQ_LOG_FORCEFREE 0x06
66#define TCP_TRK_REQ_LOG_STALE 0x07
67#define TCP_TRK_REQ_LOG_SEARCH 0x08
68
69/************************************************/
70/* Status bits we track to assure no duplicates,
71 * the bits here are not used by the code but
72 * for human representation. To check a bit we
73 * take and shift over by 1 minus the value (1-8).
74 */
75/************************************************/
76#define TCP_EI_BITS_CLIENT_FIN 0x001
77#define TCP_EI_BITS_CLIENT_RST 0x002
78#define TCP_EI_BITS_SERVER_FIN 0x004
79#define TCP_EI_BITS_SERVER_RST 0x008
80#define TCP_EI_BITS_RETRAN 0x010
81#define TCP_EI_BITS_PROGRESS 0x020
82#define TCP_EI_BITS_PRESIST_MAX 0x040
83#define TCP_EI_BITS_KEEP_MAX 0x080
84#define TCP_EI_BITS_DATA_A_CLO 0x100
85#define TCP_EI_BITS_RST_IN_FR 0x200 /* a front state reset */
86#define TCP_EI_BITS_2MS_TIMER 0x400 /* 2 MSL timer expired */
87
88#if defined(_KERNEL) || defined(_WANT_TCPCB)
89#include <sys/_callout.h>
90#include <sys/osd.h>
91
92#include <netinet/cc/cc.h>
93
94/* TCP segment queue entry */
95struct tseg_qent {
96 TAILQ_ENTRY(tseg_qent) tqe_q;
97 struct mbuf *tqe_m; /* mbuf contains packet */
98 struct mbuf *tqe_last; /* last mbuf in chain */
99 tcp_seq tqe_start; /* TCP Sequence number start */
100 int tqe_len; /* TCP segment data length */
101 uint32_t tqe_flags; /* The flags from tcp_get_flags() */
102 uint32_t tqe_mbuf_cnt; /* Count of mbuf overhead */
103};
104TAILQ_HEAD(tsegqe_head, tseg_qent);
105
106struct sackblk {
107 tcp_seq start; /* start seq no. of sack block */
108 tcp_seq end; /* end seq no. */
109};
110
111struct sackhole {
112 tcp_seq start; /* start seq no. of hole */
113 tcp_seq end; /* end seq no. */
114 tcp_seq rxmit; /* next seq. no in hole to be retransmitted */
115 TAILQ_ENTRY(sackhole) scblink; /* scoreboard linkage */
116};
117
118struct sackhint {
119 struct sackhole *nexthole;
120 int32_t sack_bytes_rexmit;
121 tcp_seq last_sack_ack; /* Most recent/largest sacked ack */
122
123 int32_t delivered_data; /* Newly acked data from last SACK */
124
125 int32_t sacked_bytes; /* Total sacked bytes reported by the
126 * receiver via sack option
127 */
128 uint32_t recover_fs; /* Flight Size at the start of Loss recovery */
129 uint32_t prr_delivered; /* Total bytes delivered using PRR */
130 uint32_t prr_out; /* Bytes sent during IN_RECOVERY */
131 int32_t hole_bytes; /* current number of bytes in scoreboard holes */
132 int32_t lost_bytes; /* number of rfc6675 IsLost() bytes */
133};
134
135#define SEGQ_EMPTY(tp) TAILQ_EMPTY(&(tp)->t_segq)
136
137STAILQ_HEAD(tcp_log_stailq, tcp_log_mem);
138
139#define TCP_TRK_TRACK_FLG_EMPTY 0x00 /* Available */
140#define TCP_TRK_TRACK_FLG_USED 0x01 /* In use */
141#define TCP_TRK_TRACK_FLG_OPEN 0x02 /* End is not valid (open range request) */
142#define TCP_TRK_TRACK_FLG_SEQV 0x04 /* We had a sendfile that touched it */
143#define TCP_TRK_TRACK_FLG_COMP 0x08 /* Sendfile as placed the last bits (range req only) */
144#define TCP_TRK_TRACK_FLG_FSND 0x10 /* First send has been done into the seq space */
145#define TCP_TRK_TRACK_FLG_LSND 0x20 /* We were able to set the Last Sent */
146#define MAX_TCP_TRK_REQ 5 /* Max we will have at once */
147
148struct tcp_sendfile_track {
149 uint64_t timestamp; /* User sent timestamp */
150 uint64_t start; /* Start of sendfile offset */
151 uint64_t end; /* End if not open-range req */
152 uint64_t localtime; /* Time we actually got the req */
153 uint64_t deadline; /* If in CU mode, deadline to delivery */
154 uint64_t first_send; /* Time of first send in the range */
155 uint64_t cspr; /* Client suggested pace rate */
156 uint64_t sent_at_fs; /* What was t_sndbytes as we begun sending */
157 uint64_t rxt_at_fs; /* What was t_snd_rxt_bytes as we begun sending */
158 uint64_t sent_at_ls; /* Sent value at the last send */
159 uint64_t rxt_at_ls; /* Retransmit value at the last send */
160 tcp_seq start_seq; /* First TCP Seq assigned */
161 tcp_seq end_seq; /* If range req last seq */
162 uint32_t flags; /* Type of request open etc */
163 uint32_t sbcc_at_s; /* When we allocate what is the sb_cc */
164 uint32_t hint_maxseg; /* Client hinted maxseg */
165 uint32_t playout_ms; /* Client playout ms */
166 uint32_t hybrid_flags; /* Hybrid flags on this request */
167};
168
169
170/*
171 * Change Query responses for a stack switch we create a structure
172 * that allows query response from the new stack to the old, if
173 * supported.
174 *
175 * There are three queries currently defined.
176 * - sendmap
177 * - timers
178 * - rack_times
179 *
180 * For the sendmap query the caller fills in the
181 * req and the req_param as the first seq (usually
182 * snd_una). When the response comes back indicating
183 * that there was data (return value 1), then the caller
184 * can build a sendmap entry based on the range and the
185 * times. The next query would then be done at the
186 * newly created sendmap_end. Repeated until sendmap_end == snd_max.
187 *
188 * Flags in sendmap_flags are defined below as well.
189 *
190 * For timers the standard PACE_TMR_XXXX flags are returned indicating
191 * a pacing timer (possibly) and one other timer. If pacing timer then
192 * the expiration timeout time in microseconds is in timer_pacing_to.
193 * And the value used with whatever timer (if a flag is set) is in
194 * timer_rxt. If no timers are running a 0 is returned and of
195 * course no flags are set in timer_hpts_flags.
196 *
197 * The rack_times are a misc collection of information that
198 * the old stack might possibly fill in. Of course its possible
199 * that an old stack may not have a piece of information. If so
200 * then setting that value to zero is advised. Setting any
201 * timestamp passed should only place a zero in it when it
202 * is unfilled. This may mean that a time is off by a micro-second
203 * but this is ok in the grand scheme of things.
204 *
205 * When switching stacks it is desireable to get as much information
206 * from the old stack to the new stack as possible. Though not always
207 * will the stack be compatible in the types of information. The
208 * init() function needs to take care when it begins changing
209 * things such as inp_flags2 and the timer units to position these
210 * changes at a point where it is unlikely they will fail after
211 * making such changes. A stack optionally can have an "undo"
212 * function
213 *
214 * To transfer information to the old stack from the new in
215 * respect to LRO and the inp_flags2, the new stack should set
216 * the inp_flags2 to what it supports. The old stack in its
217 * fini() function should call the tcp_handle_orphaned_packets()
218 * to clean up any packets. Note that a new stack should attempt
219 */
220
221/* Query types */
222#define TCP_QUERY_SENDMAP 1
223#define TCP_QUERY_TIMERS_UP 2
224#define TCP_QUERY_RACK_TIMES 3
225
226/* Flags returned in sendmap_flags */
227#define SNDMAP_ACKED 0x000001/* The remote endpoint acked this */
228#define SNDMAP_OVERMAX 0x000008/* We have more retran's then we can fit */
229#define SNDMAP_SACK_PASSED 0x000010/* A sack was done above this block */
230#define SNDMAP_HAS_FIN 0x000040/* segment is sent with fin */
231#define SNDMAP_TLP 0x000080/* segment sent as tail-loss-probe */
232#define SNDMAP_HAS_SYN 0x000800/* SYN is on this guy */
233#define SNDMAP_HAD_PUSH 0x008000/* Push was sent on original send */
234#define SNDMAP_MASK (SNDMAP_ACKED|SNDMAP_OVERMAX|SNDMAP_SACK_PASSED|SNDMAP_HAS_FIN\
235 |SNDMAP_TLP|SNDMAP_HAS_SYN|SNDMAP_HAD_PUSH)
236#define SNDMAP_NRTX 3
237
238struct tcp_query_resp {
239 int req;
240 uint32_t req_param;
241 union {
242 struct {
243 tcp_seq sendmap_start;
244 tcp_seq sendmap_end;
245 int sendmap_send_cnt;
246 uint64_t sendmap_time[SNDMAP_NRTX];
247 uint64_t sendmap_ack_arrival;
248 int sendmap_flags;
249 uint32_t sendmap_r_rtr_bytes;
250 /* If FAS is available if not 0 */
251 uint32_t sendmap_fas;
252 uint8_t sendmap_dupacks;
253 };
254 struct {
255 uint32_t timer_hpts_flags;
256 uint32_t timer_pacing_to;
257 uint32_t timer_timer_exp;
258 };
259 struct {
260 /* Timestamps and rtt's */
261 uint32_t rack_reorder_ts; /* Last uscts that reordering was seen */
262 uint32_t rack_num_dsacks; /* Num of dsacks seen */
263 uint32_t rack_rxt_last_time; /* Last time a RXT/TLP or rack tmr went off */
264 uint32_t rack_min_rtt; /* never 0 smallest rtt seen */
265 uint32_t rack_rtt; /* Last rtt used by rack */
266 uint32_t rack_tmit_time; /* The time the rtt seg was tmited */
267 uint32_t rack_time_went_idle; /* If in persist the time we went idle */
268 /* Prr data */
269 uint32_t rack_sacked;
270 uint32_t rack_holes_rxt;
271 uint32_t rack_prr_delivered;
272 uint32_t rack_prr_recovery_fs;
273 uint32_t rack_prr_out;
274 uint32_t rack_prr_sndcnt;
275 /* TLP data */
276 uint16_t rack_tlp_cnt_out; /* How many tlp's have been sent */
277 /* Various bits */
278 uint8_t rack_tlp_out; /* Is a TLP outstanding */
279 uint8_t rack_srtt_measured; /* The previous stack has measured srtt */
280 uint8_t rack_in_persist; /* Is the old stack in persists? */
281 uint8_t rack_wanted_output; /* Did the prevous stack have a want output set */
282 };
283 };
284};
285
286#define TCP_TMR_GRANULARITY_TICKS 1 /* TCP timers are in ticks (msec if hz=1000) */
287#define TCP_TMR_GRANULARITY_USEC 2 /* TCP timers are in microseconds */
288
289typedef enum {
290 TT_REXMT = 0,
291 TT_PERSIST,
292 TT_KEEP,
293 TT_2MSL,
294 TT_DELACK,
295 TT_N,
296} tt_which;
297
298typedef enum {
299 TT_PROCESSING = 0,
300 TT_PROCESSED,
301 TT_STARTING,
302 TT_STOPPING,
303} tt_what;
304
305/*
306 * Tcp control block, one per tcp connection.
307 */
308struct tcpcb {
309 struct inpcb t_inpcb; /* embedded protocol independent cb */
310#define t_start_zero t_fb
311#define t_zero_size (sizeof(struct tcpcb) - \
312 offsetof(struct tcpcb, t_start_zero))
313 struct tcp_function_block *t_fb;/* TCP function call block */
314 void *t_fb_ptr; /* Pointer to t_fb specific data */
315
316 struct callout t_callout;
317 sbintime_t t_timers[TT_N];
318 sbintime_t t_precisions[TT_N];
319
320 /* HPTS. Used by BBR and Rack stacks. See tcp_hpts.c for more info. */
321 TAILQ_ENTRY(tcpcb) t_hpts; /* linkage to HPTS ring */
322 STAILQ_HEAD(, mbuf) t_inqueue; /* HPTS input packets queue */
323 uint32_t t_hpts_request; /* Current hpts request, zero if
324 * fits in the pacing window. */
325 uint32_t t_hpts_slot; /* HPTS wheel slot this tcb is. */
326 uint32_t t_hpts_drop_reas; /* Reason we are dropping the pcb. */
327 uint32_t t_hpts_gencnt;
328 uint16_t t_hpts_cpu; /* CPU chosen by hpts_cpuid(). */
329 uint16_t t_lro_cpu; /* CPU derived from LRO. */
330#define HPTS_CPU_NONE ((uint16_t)-1)
331 enum {
332 IHPTS_NONE = 0,
333 IHPTS_ONQUEUE,
334 IHPTS_MOVING,
335 } t_in_hpts; /* Is it linked into HPTS? */
336
337 uint32_t t_maxseg:24, /* maximum segment size */
338 _t_logstate:8; /* State of "black box" logging */
339 uint32_t t_port:16, /* Tunneling (over udp) port */
340 t_state:4, /* state of this connection */
341 t_idle_reduce : 1,
342 t_delayed_ack: 7, /* Delayed ack variable */
343 t_fin_is_rst: 1, /* Are fin's treated as resets */
344 t_log_state_set: 1,
345 bits_spare : 2;
346 u_int t_flags;
347 tcp_seq snd_una; /* sent but unacknowledged */
348 tcp_seq snd_max; /* highest sequence number sent;
349 * used to recognize retransmits
350 */
351 tcp_seq snd_nxt; /* send next */
352 tcp_seq snd_up; /* send urgent pointer */
353 uint32_t snd_wnd; /* send window */
354 uint32_t snd_cwnd; /* congestion-controlled window */
355 uint32_t ts_offset; /* our timestamp offset */
356 uint32_t rfbuf_ts; /* recv buffer autoscaling timestamp */
357 int rcv_numsacks; /* # distinct sack blks present */
358 u_int t_tsomax; /* TSO total burst length limit */
359 u_int t_tsomaxsegcount; /* TSO maximum segment count */
360 u_int t_tsomaxsegsize; /* TSO maximum segment size in bytes */
361 tcp_seq rcv_nxt; /* receive next */
362 tcp_seq rcv_adv; /* advertised window */
363 uint32_t rcv_wnd; /* receive window */
364 u_int t_flags2; /* More tcpcb flags storage */
365 int t_srtt; /* smoothed round-trip time */
366 int t_rttvar; /* variance in round-trip time */
367 uint32_t ts_recent; /* timestamp echo data */
368 u_char snd_scale; /* window scaling for send window */
369 u_char rcv_scale; /* window scaling for recv window */
370 u_char snd_limited; /* segments limited transmitted */
371 u_char request_r_scale; /* pending window scaling */
372 tcp_seq last_ack_sent;
373 u_int t_rcvtime; /* inactivity time */
374 tcp_seq rcv_up; /* receive urgent pointer */
375 int t_segqlen; /* segment reassembly queue length */
376 uint32_t t_segqmbuflen; /* total reassembly queue byte length */
377 struct tsegqe_head t_segq; /* segment reassembly queue */
378 uint32_t snd_ssthresh; /* snd_cwnd size threshold for
379 * for slow start exponential to
380 * linear switch
381 */
382 tcp_seq snd_wl1; /* window update seg seq number */
383 tcp_seq snd_wl2; /* window update seg ack number */
384
385 tcp_seq irs; /* initial receive sequence number */
386 tcp_seq iss; /* initial send sequence number */
387 u_int t_acktime; /* RACK and BBR incoming new data was acked */
388 u_int t_sndtime; /* time last data was sent */
389 u_int ts_recent_age; /* when last updated */
390 tcp_seq snd_recover; /* for use in NewReno Fast Recovery */
391 char t_oobflags; /* have some */
392 char t_iobc; /* input character */
393 uint8_t t_nic_ktls_xmit:1, /* active nic ktls xmit sessions */
394 t_nic_ktls_xmit_dis:1, /* disabled nic xmit ktls? */
395 t_nic_ktls_spare:6; /* spare nic ktls */
396 int t_rxtcur; /* current retransmit value (ticks) */
397
398 int t_rxtshift; /* log(2) of rexmt exp. backoff */
399 u_int t_rtttime; /* RTT measurement start time */
400
401 tcp_seq t_rtseq; /* sequence number being timed */
402 u_int t_starttime; /* time connection was established */
403 u_int t_fbyte_in; /* ticks time first byte queued in */
404 u_int t_fbyte_out; /* ticks time first byte queued out */
405
406 u_int t_pmtud_saved_maxseg; /* pre-blackhole MSS */
407 int t_blackhole_enter; /* when to enter blackhole detection */
408 int t_blackhole_exit; /* when to exit blackhole detection */
409 u_int t_rttmin; /* minimum rtt allowed */
410
411 int t_softerror; /* possible error not yet reported */
412 uint32_t max_sndwnd; /* largest window peer has offered */
413 uint32_t snd_cwnd_prev; /* cwnd prior to retransmit */
414 uint32_t snd_ssthresh_prev; /* ssthresh prior to retransmit */
415 tcp_seq snd_recover_prev; /* snd_recover prior to retransmit */
416 int t_sndzerowin; /* zero-window updates sent */
417 int snd_numholes; /* number of holes seen by sender */
418 u_int t_badrxtwin; /* window for retransmit recovery */
419 TAILQ_HEAD(sackhole_head, sackhole) snd_holes;
420 /* SACK scoreboard (sorted) */
421 tcp_seq snd_fack; /* last seq number(+1) sack'd by rcv'r*/
422 struct sackblk sackblks[MAX_SACK_BLKS]; /* seq nos. of sack blocks */
423 struct sackhint sackhint; /* SACK scoreboard hint */
424 int t_rttlow; /* smallest observerved RTT */
425 int rfbuf_cnt; /* recv buffer autoscaling byte count */
426 struct toedev *tod; /* toedev handling this connection */
427 int t_sndrexmitpack; /* retransmit packets sent */
428 int t_rcvoopack; /* out-of-order packets received */
429 void *t_toe; /* TOE pcb pointer */
430 struct cc_algo *t_cc; /* congestion control algorithm */
431 struct cc_var t_ccv; /* congestion control specific vars */
432 int t_bytes_acked; /* # bytes acked during current RTT */
433 u_int t_maxunacktime;
434 u_int t_keepinit; /* time to establish connection */
435 u_int t_keepidle; /* time before keepalive probes begin */
436 u_int t_keepintvl; /* interval between keepalives */
437 u_int t_keepcnt; /* number of keepalives before close */
438 int t_dupacks; /* consecutive dup acks recd */
439 int t_lognum; /* Number of log entries */
440 int t_loglimit; /* Maximum number of log entries */
441 uint32_t t_rcep; /* Number of received CE marked pkts */
442 uint32_t t_scep; /* Synced number of delivered CE pkts */
443 int64_t t_pacing_rate; /* bytes / sec, -1 => unlimited */
444 struct tcp_log_stailq t_logs; /* Log buffer */
445 struct tcp_log_id_node *t_lin;
446 struct tcp_log_id_bucket *t_lib;
447 const char *t_output_caller; /* Function that called tcp_output */
448 struct statsblob *t_stats; /* Per-connection stats */
449 /* Should these be a pointer to the arrays or an array? */
450 uint32_t t_logsn; /* Log "serial number" */
451 uint32_t gput_ts; /* Time goodput measurement started */
452 tcp_seq gput_seq; /* Outbound measurement seq */
453 tcp_seq gput_ack; /* Inbound measurement ack */
454 int32_t t_stats_gput_prev; /* XXXLAS: Prev gput measurement */
455 uint32_t t_sndtlppack; /* tail loss probe packets sent */
456 uint64_t t_sndtlpbyte; /* total tail loss probe bytes sent */
457 uint64_t t_sndbytes; /* total bytes sent */
458 uint64_t t_snd_rxt_bytes; /* total bytes retransmitted */
459 uint32_t t_dsack_bytes; /* dsack bytes received */
460 uint32_t t_dsack_tlp_bytes; /* dsack bytes received for TLPs sent */
461 uint32_t t_dsack_pack; /* dsack packets we have eceived */
462 uint8_t t_tmr_granularity; /* Granularity of all timers srtt etc */
463 uint8_t t_rttupdated; /* number of times rtt sampled */
464 /* TCP Fast Open */
465 uint8_t t_tfo_client_cookie_len; /* TFO client cookie length */
466 uint32_t t_end_info_status; /* Status flag of end info */
467 sbintime_t t_challenge_ack_end; /* End of the challenge ack epoch */
468 uint32_t t_challenge_ack_cnt; /* Number of challenge ACKs sent in
469 * current epoch
470 */
471
472 unsigned int *t_tfo_pending; /* TFO server pending counter */
473 union {
474 uint8_t client[TCP_FASTOPEN_MAX_COOKIE_LEN];
475 uint64_t server;
476 } t_tfo_cookie; /* TCP Fast Open cookie to send */
477 union {
478 uint8_t t_end_info_bytes[TCP_END_BYTE_INFO];
479 uint64_t t_end_info;
480 };
481 struct osd t_osd; /* storage for Khelp module data */
482 uint8_t _t_logpoint; /* Used when a BB log points is enabled */
483 /*
484 * Keep all #ifdef'ed components at the end of the structure!
485 * This is important to minimize problems when compiling modules
486 * using this structure from within the modules' directory.
487 */
488#ifdef TCP_REQUEST_TRK
489 /* Response tracking addons. */
490 uint8_t t_tcpreq_req; /* Request count */
491 uint8_t t_tcpreq_open; /* Number of open range requests */
492 uint8_t t_tcpreq_closed; /* Number of closed range requests */
493 uint32_t tcp_hybrid_start; /* Num of times we started hybrid pacing */
494 uint32_t tcp_hybrid_stop; /* Num of times we stopped hybrid pacing */
495 uint32_t tcp_hybrid_error; /* Num of times we failed to start hybrid pacing */
496 struct tcp_sendfile_track t_tcpreq_info[MAX_TCP_TRK_REQ];
497#endif
498#ifdef TCP_ACCOUNTING
499 uint64_t tcp_cnt_counters[TCP_NUM_CNT_COUNTERS];
500 uint64_t tcp_proc_time[TCP_NUM_CNT_COUNTERS];
501#endif
502};
503#endif /* _KERNEL || _WANT_TCPCB */
504
505#ifdef _KERNEL
506struct tcptemp {
507 u_char tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */
508 struct tcphdr tt_t;
509};
510
511/* SACK scoreboard update status */
512typedef enum {
513 SACK_NOCHANGE = 0,
514 SACK_CHANGE,
515 SACK_NEWLOSS
516} sackstatus_t;
517
518/* Enable TCP/UDP tunneling port */
519#define TCP_TUNNELING_PORT_MIN 0
520#define TCP_TUNNELING_PORT_MAX 65535
521#define TCP_TUNNELING_PORT_DEFAULT 0
522
523/* Enable TCP/UDP tunneling port */
524#define TCP_TUNNELING_OVERHEAD_MIN sizeof(struct udphdr)
525#define TCP_TUNNELING_OVERHEAD_MAX 1024
526#define TCP_TUNNELING_OVERHEAD_DEFAULT TCP_TUNNELING_OVERHEAD_MIN
527
528/* Minimum map entries limit value, if set */
529#define TCP_MIN_MAP_ENTRIES_LIMIT 128
530
531/* Flags for tcp functions */
532#define TCP_FUNC_BEING_REMOVED 0x01 /* Can no longer be referenced */
533#define TCP_FUNC_OUTPUT_CANDROP 0x02 /* tfb_tcp_output may ask tcp_drop */
534#define TCP_FUNC_DEFAULT_OK 0x04 /* Can be used as default */
535
536/**
537 * tfb_tcp_handoff_ok is a mandatory function allowing
538 * to query a stack, if it can take over a tcpcb.
539 * You return 0 to say you can take over and run your stack,
540 * you return non-zero (an error number) to say no you can't.
541 *
542 * tfb_tcp_fb_init is used to allow the new stack to
543 * setup its control block. Among the things it must
544 * do is:
545 * a) Make sure that the inp_flags2 is setup correctly
546 * for LRO. There are two flags that the previous
547 * stack may have set INP_MBUF_ACKCMP and
548 * INP_SUPPORTS_MBUFQ. If the new stack does not
549 * support these it *should* clear the flags.
550 * b) Make sure that the timers are in the proper
551 * granularity that the stack wants. The stack
552 * should check the t_tmr_granularity field. Currently
553 * there are two values that it may hold
554 * TCP_TMR_GRANULARITY_TICKS and TCP_TMR_GRANULARITY_USEC.
555 * Use the functions tcp_timer_convert(tp, granularity);
556 * to move the timers to the correct format for your stack.
557 *
558 * The new stack may also optionally query the tfb_chg_query
559 * function if the old stack has one. The new stack may ask
560 * for one of three entries and can also state to the old
561 * stack its support for the INP_MBUF_ACKCMP and
562 * INP_SUPPORTS_MBUFQ. This is important since if there are
563 * queued ack's without that statement the old stack will
564 * be forced to discard the queued acks. The requests that
565 * can be made for information by the new stacks are:
566 *
567 * Note also that the tfb_tcp_fb_init() when called can
568 * determine if a query is needed by looking at the
569 * value passed in the ptr. The ptr is designed to be
570 * set in with any allocated memory, but the address
571 * of the condtion (ptr == &tp->t_fb_ptr) will be
572 * true if this is not a stack switch but the initial
573 * setup of a tcb (which means no query would be needed).
574 * If, however, the value is not t_fb_ptr, then the caller
575 * is in the middle of a stack switch and is the new stack.
576 * A query would be appropriate (if the new stack support
577 * the query mechanism).
578 *
579 * TCP_QUERY_SENDMAP - Query of outstanding data.
580 * TCP_QUERY_TIMERS_UP - Query about running timers.
581 * TCP_SUPPORTED_LRO - Declaration in req_param of
582 * the inp_flags2 supported by
583 * the new stack.
584 * TCP_QUERY_RACK_TIMES - Enquire about various timestamps
585 * and states the old stack may be in.
586 *
587 * tfb_tcp_fb_fini is changed to add a flag to tell
588 * the old stack if the tcb is being destroyed or
589 * not. A one in the flag means the TCB is being
590 * destroyed, a zero indicates its transitioning to
591 * another stack (via socket option). The
592 * tfb_tcp_fb_fini() function itself should not change timers
593 * or inp_flags2 (the tfb_tcp_fb_init() must do that). However
594 * if the old stack supports the LRO mbuf queuing, and the new
595 * stack does not communicate via chg messages that it too does,
596 * it must assume it does not and free any queued mbufs.
597 *
598 */
599struct tcp_function_block {
600 char tfb_tcp_block_name[TCP_FUNCTION_NAME_LEN_MAX];
601 int (*tfb_tcp_output)(struct tcpcb *);
602 void (*tfb_tcp_do_segment)(struct tcpcb *, struct mbuf *,
603 struct tcphdr *, int, int, uint8_t);
604 int (*tfb_do_segment_nounlock)(struct tcpcb *, struct mbuf *,
605 struct tcphdr *, int, int, uint8_t, int, struct timeval *);
606 int (*tfb_do_queued_segments)(struct tcpcb *, int);
607 int (*tfb_tcp_ctloutput)(struct tcpcb *, struct sockopt *);
608 /* Optional memory allocation/free routine */
609 int (*tfb_tcp_fb_init)(struct tcpcb *, void **);
610 void (*tfb_tcp_fb_fini)(struct tcpcb *, int);
611 /* Optional timers, must define all if you define one */
612 int (*tfb_tcp_timer_stop_all)(struct tcpcb *);
613 void (*tfb_tcp_rexmit_tmr)(struct tcpcb *);
614 int (*tfb_tcp_handoff_ok)(struct tcpcb *);
615 void (*tfb_tcp_mtu_chg)(struct tcpcb *tp);
616 int (*tfb_pru_options)(struct tcpcb *, int);
617 void (*tfb_hwtls_change)(struct tcpcb *, int);
618 int (*tfb_chg_query)(struct tcpcb *, struct tcp_query_resp *);
619 void (*tfb_switch_failed)(struct tcpcb *);
620 bool (*tfb_early_wake_check)(struct tcpcb *);
621 int (*tfb_compute_pipe)(struct tcpcb *tp);
622 int (*tfb_stack_info)(struct tcpcb *tp, struct stack_specific_info *);
623 void (*tfb_inherit)(struct tcpcb *tp, struct inpcb *h_inp);
624 volatile uint32_t tfb_refcnt;
625 uint32_t tfb_flags;
626 uint8_t tfb_id;
627};
628
629/* Maximum number of names each TCP function block can be registered with. */
630#define TCP_FUNCTION_NAME_NUM_MAX 8
631
632struct tcp_function {
633 TAILQ_ENTRY(tcp_function) tf_next;
634 char tf_name[TCP_FUNCTION_NAME_LEN_MAX];
635 struct tcp_function_block *tf_fb;
636};
637
638TAILQ_HEAD(tcp_funchead, tcp_function);
639
640struct tcpcb * tcp_drop(struct tcpcb *, int);
641
642#ifdef _NETINET_IN_PCB_H_
643#define intotcpcb(inp) __containerof((inp), struct tcpcb, t_inpcb)
644#define sototcpcb(so) intotcpcb(sotoinpcb(so))
645#define tptoinpcb(tp) (&(tp)->t_inpcb)
646#define tptosocket(tp) (tp)->t_inpcb.inp_socket
647
648/*
649 * tcp_output()
650 * Handles tcp_drop request from advanced stacks and reports that inpcb is
651 * gone with negative return code.
652 * Drop in replacement for the default stack.
653 */
654static inline int
655tcp_output(struct tcpcb *tp)
656{
657 struct inpcb *inp = tptoinpcb(tp);
658 int rv;
659
660 INP_WLOCK_ASSERT(inp);
661
662 rv = tp->t_fb->tfb_tcp_output(tp);
663 if (rv < 0) {
664 KASSERT(tp->t_fb->tfb_flags & TCP_FUNC_OUTPUT_CANDROP,
665 ("TCP stack %s requested tcp_drop(%p)",
666 tp->t_fb->tfb_tcp_block_name, tp));
667 tp = tcp_drop(tp, -rv);
668 if (tp)
669 INP_WUNLOCK(inp);
670 }
671
672 return (rv);
673}
674
675/*
676 * tcp_output_unlock()
677 * Always returns unlocked, handles drop request from advanced stacks.
678 * Always returns positive error code.
679 */
680static inline int
681tcp_output_unlock(struct tcpcb *tp)
682{
683 struct inpcb *inp = tptoinpcb(tp);
684 int rv;
685
686 INP_WLOCK_ASSERT(inp);
687
688 rv = tp->t_fb->tfb_tcp_output(tp);
689 if (rv < 0) {
690 KASSERT(tp->t_fb->tfb_flags & TCP_FUNC_OUTPUT_CANDROP,
691 ("TCP stack %s requested tcp_drop(%p)",
692 tp->t_fb->tfb_tcp_block_name, tp));
693 rv = -rv;
694 tp = tcp_drop(tp, rv);
695 if (tp)
696 INP_WUNLOCK(inp);
697 } else
698 INP_WUNLOCK(inp);
699
700 return (rv);
701}
702
703/*
704 * tcp_output_nodrop()
705 * Always returns locked. It is caller's responsibility to run tcp_drop()!
706 * Useful in syscall implementations, when we want to perform some logging
707 * and/or tracing with tcpcb before calling tcp_drop(). To be used with
708 * tcp_unlock_or_drop() later.
709 *
710 * XXXGL: maybe don't allow stacks to return a drop request at certain
711 * TCP states? Why would it do in connect(2)? In recv(2)?
712 */
713static inline int
714tcp_output_nodrop(struct tcpcb *tp)
715{
716 int rv;
717
718 INP_WLOCK_ASSERT(tptoinpcb(tp));
719
720 rv = tp->t_fb->tfb_tcp_output(tp);
721 KASSERT(rv >= 0 || tp->t_fb->tfb_flags & TCP_FUNC_OUTPUT_CANDROP,
722 ("TCP stack %s requested tcp_drop(%p)",
723 tp->t_fb->tfb_tcp_block_name, tp));
724 return (rv);
725}
726
727/*
728 * tcp_unlock_or_drop()
729 * Handle return code from tfb_tcp_output() after we have logged/traced,
730 * to be used with tcp_output_nodrop().
731 */
732static inline int
733tcp_unlock_or_drop(struct tcpcb *tp, int tcp_output_retval)
734{
735 struct inpcb *inp = tptoinpcb(tp);
736
737 INP_WLOCK_ASSERT(inp);
738
739 if (tcp_output_retval < 0) {
740 tcp_output_retval = -tcp_output_retval;
741 if (tcp_drop(tp, tcp_output_retval) != NULL)
742 INP_WUNLOCK(inp);
743 } else
744 INP_WUNLOCK(inp);
745
746 return (tcp_output_retval);
747}
748#endif /* _NETINET_IN_PCB_H_ */
749
750static int inline
751tcp_packets_this_ack(struct tcpcb *tp, tcp_seq ack)
752{
753 return ((ack - tp->snd_una) / tp->t_maxseg +
754 ((((ack - tp->snd_una) % tp->t_maxseg) != 0) ? 1 : 0));
755}
756#endif /* _KERNEL */
757
758/*
759 * Flags and utility macros for the t_flags field.
760 */
761#define TF_ACKNOW 0x00000001 /* ack peer immediately */
762#define TF_DELACK 0x00000002 /* ack, but try to delay it */
763#define TF_NODELAY 0x00000004 /* don't delay packets to coalesce */
764#define TF_NOOPT 0x00000008 /* don't use tcp options */
765#define TF_SENTFIN 0x00000010 /* have sent FIN */
766#define TF_REQ_SCALE 0x00000020 /* have/will request window scaling */
767#define TF_RCVD_SCALE 0x00000040 /* other side has requested scaling */
768#define TF_REQ_TSTMP 0x00000080 /* have/will request timestamps */
769#define TF_RCVD_TSTMP 0x00000100 /* a timestamp was received in SYN */
770#define TF_SACK_PERMIT 0x00000200 /* other side said I could SACK */
771#define TF_NEEDSYN 0x00000400 /* send SYN (implicit state) */
772#define TF_NEEDFIN 0x00000800 /* send FIN (implicit state) */
773#define TF_NOPUSH 0x00001000 /* don't push */
774#define TF_PREVVALID 0x00002000 /* saved values for bad rxmit valid
775 * Note: accessing and restoring from
776 * these may only be done in the 1st
777 * RTO recovery round (t_rxtshift == 1)
778 */
779#define TF_WAKESOR 0x00004000 /* wake up receive socket */
780#define TF_GPUTINPROG 0x00008000 /* Goodput measurement in progress */
781#define TF_MORETOCOME 0x00010000 /* More data to be appended to sock */
782#define TF_SONOTCONN 0x00020000 /* needs soisconnected() on ESTAB */
783#define TF_LASTIDLE 0x00040000 /* connection was previously idle */
784#define TF_RXWIN0SENT 0x00080000 /* sent a receiver win 0 in response */
785#define TF_FASTRECOVERY 0x00100000 /* in NewReno Fast Recovery */
786#define TF_WASFRECOVERY 0x00200000 /* was in NewReno Fast Recovery */
787#define TF_SIGNATURE 0x00400000 /* require MD5 digests (RFC2385) */
788#define TF_FORCEDATA 0x00800000 /* force out a byte */
789#define TF_TSO 0x01000000 /* TSO enabled on this connection */
790#define TF_TOE 0x02000000 /* this connection is offloaded */
791#define TF_CLOSED 0x04000000 /* close(2) called on socket */
792#define TF_SENTSYN 0x08000000 /* At least one syn has been sent */
793#define TF_LRD 0x10000000 /* Lost Retransmission Detection */
794#define TF_CONGRECOVERY 0x20000000 /* congestion recovery mode */
795#define TF_WASCRECOVERY 0x40000000 /* was in congestion recovery */
796#define TF_FASTOPEN 0x80000000 /* TCP Fast Open indication */
797
798#define IN_FASTRECOVERY(t_flags) (t_flags & TF_FASTRECOVERY)
799#define ENTER_FASTRECOVERY(t_flags) t_flags |= TF_FASTRECOVERY
800#define EXIT_FASTRECOVERY(t_flags) t_flags &= ~TF_FASTRECOVERY
801
802#define IN_CONGRECOVERY(t_flags) (t_flags & TF_CONGRECOVERY)
803#define ENTER_CONGRECOVERY(t_flags) t_flags |= TF_CONGRECOVERY
804#define EXIT_CONGRECOVERY(t_flags) t_flags &= ~TF_CONGRECOVERY
805
806#define IN_RECOVERY(t_flags) (t_flags & (TF_CONGRECOVERY | TF_FASTRECOVERY))
807#define ENTER_RECOVERY(t_flags) t_flags |= (TF_CONGRECOVERY | TF_FASTRECOVERY)
808#define EXIT_RECOVERY(t_flags) t_flags &= ~(TF_CONGRECOVERY | TF_FASTRECOVERY)
809
810#define BYTES_THIS_ACK(tp, th) (th->th_ack - tp->snd_una)
811
812/*
813 * Flags for the t_oobflags field.
814 */
815#define TCPOOB_HAVEDATA 0x01
816#define TCPOOB_HADDATA 0x02
817
818/*
819 * Flags for the extended TCP flags field, t_flags2
820 */
821#define TF2_PLPMTU_BLACKHOLE 0x00000001 /* Possible PLPMTUD Black Hole. */
822#define TF2_PLPMTU_PMTUD 0x00000002 /* Allowed to attempt PLPMTUD. */
823#define TF2_PLPMTU_MAXSEGSNT 0x00000004 /* Last seg sent was full seg. */
824#define TF2_LOG_AUTO 0x00000008 /* Session is auto-logging. */
825#define TF2_DROP_AF_DATA 0x00000010 /* Drop after all data ack'd */
826#define TF2_ECN_PERMIT 0x00000020 /* connection ECN-ready */
827#define TF2_ECN_SND_CWR 0x00000040 /* ECN CWR in queue */
828#define TF2_ECN_SND_ECE 0x00000080 /* ECN ECE in queue */
829#define TF2_ACE_PERMIT 0x00000100 /* Accurate ECN mode */
830#define TF2_HPTS_CPU_SET 0x00000200 /* t_hpts_cpu is not random */
831#define TF2_FBYTES_COMPLETE 0x00000400 /* We have first bytes in and out */
832#define TF2_ECN_USE_ECT1 0x00000800 /* Use ECT(1) marking on session */
833#define TF2_TCP_ACCOUNTING 0x00001000 /* Do TCP accounting */
834#define TF2_HPTS_CALLS 0x00002000 /* tcp_output() called via HPTS */
835#define TF2_MBUF_L_ACKS 0x00004000 /* large mbufs for ack compression */
836#define TF2_MBUF_ACKCMP 0x00008000 /* mbuf ack compression ok */
837#define TF2_SUPPORTS_MBUFQ 0x00010000 /* Supports the mbuf queue method */
838#define TF2_MBUF_QUEUE_READY 0x00020000 /* Inputs can be queued */
839#define TF2_DONT_SACK_QUEUE 0x00040000 /* Don't wake on sack */
840#define TF2_CANNOT_DO_ECN 0x00080000 /* The stack does not do ECN */
841#define TF2_PROC_SACK_PROHIBIT 0x00100000 /* Due to small MSS size do not process sack's */
842#define TF2_IPSEC_TSO 0x00200000 /* IPSEC + TSO supported */
843#define TF2_NO_ISS_CHECK 0x00400000 /* Don't check SEG.ACK against ISS */
844
845/*
846 * Structure to hold TCP options that are only used during segment
847 * processing (in tcp_input), but not held in the tcpcb.
848 * It's basically used to reduce the number of parameters
849 * to tcp_dooptions and tcp_addoptions.
850 * The binary order of the to_flags is relevant for packing of the
851 * options in tcp_addoptions.
852 */
853struct tcpopt {
854 u_int32_t to_flags; /* which options are present */
855#define TOF_MSS 0x0001 /* maximum segment size */
856#define TOF_SCALE 0x0002 /* window scaling */
857#define TOF_SACKPERM 0x0004 /* SACK permitted */
858#define TOF_TS 0x0010 /* timestamp */
859#define TOF_SIGNATURE 0x0040 /* TCP-MD5 signature option (RFC2385) */
860#define TOF_SACK 0x0080 /* Peer sent SACK option */
861#define TOF_FASTOPEN 0x0100 /* TCP Fast Open (TFO) cookie */
862#define TOF_MAXOPT 0x0200
863 u_int32_t to_tsval; /* new timestamp */
864 u_int32_t to_tsecr; /* reflected timestamp */
865 u_char *to_sacks; /* pointer to the first SACK blocks */
866 u_char *to_signature; /* pointer to the TCP-MD5 signature */
867 u_int8_t *to_tfo_cookie; /* pointer to the TFO cookie */
868 u_int16_t to_mss; /* maximum segment size */
869 u_int8_t to_wscale; /* window scaling */
870 u_int8_t to_nsacks; /* number of SACK blocks */
871 u_int8_t to_tfo_len; /* TFO cookie length */
872 u_int32_t to_spare; /* UTO */
873};
874
875/*
876 * Flags for tcp_dooptions.
877 */
878#define TO_SYN 0x01 /* parse SYN-only options */
879
880struct hc_metrics_lite { /* must stay in sync with hc_metrics */
881 uint32_t hc_mtu; /* MTU for this path */
882 uint32_t hc_ssthresh; /* outbound gateway buffer limit */
883 uint32_t hc_rtt; /* estimated round trip time */
884 uint32_t hc_rttvar; /* estimated rtt variance */
885 uint32_t hc_cwnd; /* congestion window */
886 uint32_t hc_sendpipe; /* outbound delay-bandwidth product */
887 uint32_t hc_recvpipe; /* inbound delay-bandwidth product */
888};
889
890#ifndef _NETINET_IN_PCB_H_
891struct in_conninfo;
892#endif /* _NETINET_IN_PCB_H_ */
893
894/*
895 * The smoothed round-trip time and estimated variance
896 * are stored as fixed point numbers scaled by the values below.
897 * For convenience, these scales are also used in smoothing the average
898 * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
899 * With these scales, srtt has 3 bits to the right of the binary point,
900 * and thus an "ALPHA" of 0.875. rttvar has 2 bits to the right of the
901 * binary point, and is smoothed with an ALPHA of 0.75.
902 */
903#define TCP_RTT_SCALE 32 /* multiplier for srtt; 5 bits frac. */
904#define TCP_RTT_SHIFT 5 /* shift for srtt; 5 bits frac. */
905#define TCP_RTTVAR_SCALE 16 /* multiplier for rttvar; 4 bits */
906#define TCP_RTTVAR_SHIFT 4 /* shift for rttvar; 4 bits */
907#define TCP_DELTA_SHIFT 2 /* see tcp_input.c */
908
909/*
910 * The initial retransmission should happen at rtt + 4 * rttvar.
911 * Because of the way we do the smoothing, srtt and rttvar
912 * will each average +1/2 tick of bias. When we compute
913 * the retransmit timer, we want 1/2 tick of rounding and
914 * 1 extra tick because of +-1/2 tick uncertainty in the
915 * firing of the timer. The bias will give us exactly the
916 * 1.5 tick we need. But, because the bias is
917 * statistical, we have to test that we don't drop below
918 * the minimum feasible timer (which is 2 ticks).
919 * This version of the macro adapted from a paper by Lawrence
920 * Brakmo and Larry Peterson which outlines a problem caused
921 * by insufficient precision in the original implementation,
922 * which results in inappropriately large RTO values for very
923 * fast networks.
924 */
925#define TCP_REXMTVAL(tp) \
926 max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT)) \
927 + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
928
929/*
930 * Global (per-VNET) TCP statistics. The below structure represents what we
931 * export to the userland, but in the kernel we have an array of counter_u64_t
932 * with as many elements as there are members in the structure. The counters
933 * shall be increased by TCPSTAT_INC() or KMOD_TCPSTAT_INC(). Adding a new
934 * counter also requires adding corresponding SDT probes into in_kdtrace.h and
935 * into in_kdtrace.c.
936 */
937struct tcpstat {
938 uint64_t tcps_connattempt; /* connections initiated */
939 uint64_t tcps_accepts; /* connections accepted */
940 uint64_t tcps_connects; /* connections established */
941 uint64_t tcps_drops; /* connections dropped */
942 uint64_t tcps_conndrops; /* embryonic connections dropped */
943 uint64_t tcps_minmssdrops; /* average minmss too low drops */
944 uint64_t tcps_closed; /* conn. closed (includes drops) */
945 uint64_t tcps_segstimed; /* segs where we tried to get rtt */
946 uint64_t tcps_rttupdated; /* times we succeeded */
947 uint64_t tcps_delack; /* delayed acks sent */
948 uint64_t tcps_timeoutdrop; /* conn. dropped in rxmt timeout */
949 uint64_t tcps_rexmttimeo; /* retransmit timeouts */
950 uint64_t tcps_persisttimeo; /* persist timeouts */
951 uint64_t tcps_keeptimeo; /* keepalive timeouts */
952 uint64_t tcps_keepprobe; /* keepalive probes sent */
953 uint64_t tcps_keepdrops; /* connections dropped in keepalive */
954 uint64_t tcps_progdrops; /* drops due to no progress */
955
956 uint64_t tcps_sndtotal; /* total packets sent */
957 uint64_t tcps_sndpack; /* data packets sent */
958 uint64_t tcps_sndbyte; /* data bytes sent */
959 uint64_t tcps_sndrexmitpack; /* data packets retransmitted */
960 uint64_t tcps_sndrexmitbyte; /* data bytes retransmitted */
961 uint64_t tcps_sndrexmitbad; /* unnecessary packet retransmissions */
962 uint64_t tcps_sndacks; /* ack-only packets sent */
963 uint64_t tcps_sndprobe; /* window probes sent */
964 uint64_t tcps_sndurg; /* packets sent with URG only */
965 uint64_t tcps_sndwinup; /* window update-only packets sent */
966 uint64_t tcps_sndctrl; /* control (SYN|FIN|RST) packets sent */
967
968 uint64_t tcps_rcvtotal; /* total packets received */
969 uint64_t tcps_rcvpack; /* packets received in sequence */
970 uint64_t tcps_rcvbyte; /* bytes received in sequence */
971 uint64_t tcps_rcvbadsum; /* packets received with ccksum errs */
972 uint64_t tcps_rcvbadoff; /* packets received with bad offset */
973 uint64_t tcps_rcvreassfull; /* packets dropped for no reass space */
974 uint64_t tcps_rcvshort; /* packets received too short */
975 uint64_t tcps_rcvduppack; /* duplicate-only packets received */
976 uint64_t tcps_rcvdupbyte; /* duplicate-only bytes received */
977 uint64_t tcps_rcvpartduppack; /* packets with some duplicate data */
978 uint64_t tcps_rcvpartdupbyte; /* dup. bytes in part-dup. packets */
979 uint64_t tcps_rcvoopack; /* out-of-order packets received */
980 uint64_t tcps_rcvoobyte; /* out-of-order bytes received */
981 uint64_t tcps_rcvpackafterwin; /* packets with data after window */
982 uint64_t tcps_rcvbyteafterwin; /* bytes rcvd after window */
983 uint64_t tcps_rcvafterclose; /* packets rcvd after "close" */
984 uint64_t tcps_rcvwinprobe; /* rcvd window probe packets */
985 uint64_t tcps_rcvdupack; /* rcvd duplicate acks */
986 uint64_t tcps_rcvacktoomuch; /* rcvd acks for unsent data */
987 uint64_t tcps_rcvackpack; /* rcvd ack packets */
988 uint64_t tcps_rcvackbyte; /* bytes acked by rcvd acks */
989 uint64_t tcps_rcvwinupd; /* rcvd window update packets */
990 uint64_t tcps_pawsdrop; /* segments dropped due to PAWS */
991 uint64_t tcps_predack; /* times hdr predict ok for acks */
992 uint64_t tcps_preddat; /* times hdr predict ok for data pkts */
993 uint64_t tcps_pcbcachemiss;
994 uint64_t tcps_cachedrtt; /* times cached RTT in route updated */
995 uint64_t tcps_cachedrttvar; /* times cached rttvar updated */
996 uint64_t tcps_cachedssthresh; /* times cached ssthresh updated */
997 uint64_t tcps_usedrtt; /* times RTT initialized from route */
998 uint64_t tcps_usedrttvar; /* times RTTVAR initialized from rt */
999 uint64_t tcps_usedssthresh; /* times ssthresh initialized from rt*/
1000 uint64_t tcps_persistdrop; /* timeout in persist state */
1001 uint64_t tcps_badsyn; /* bogus SYN, e.g. premature ACK */
1002 uint64_t tcps_mturesent; /* resends due to MTU discovery */
1003 uint64_t tcps_listendrop; /* listen queue overflows */
1004 uint64_t tcps_badrst; /* ignored RSTs in the window */
1005
1006 uint64_t tcps_sc_added; /* entry added to syncache */
1007 uint64_t tcps_sc_retransmitted; /* syncache entry was retransmitted */
1008 uint64_t tcps_sc_dupsyn; /* duplicate SYN packet */
1009 uint64_t tcps_sc_dropped; /* could not reply to packet */
1010 uint64_t tcps_sc_completed; /* successful extraction of entry */
1011 uint64_t tcps_sc_bucketoverflow;/* syncache per-bucket limit hit */
1012 uint64_t tcps_sc_cacheoverflow; /* syncache cache limit hit */
1013 uint64_t tcps_sc_reset; /* RST removed entry from syncache */
1014 uint64_t tcps_sc_stale; /* timed out or listen socket gone */
1015 uint64_t tcps_sc_aborted; /* syncache entry aborted */
1016 uint64_t tcps_sc_badack; /* removed due to bad ACK */
1017 uint64_t tcps_sc_unreach; /* ICMP unreachable received */
1018 uint64_t tcps_sc_zonefail; /* zalloc() failed */
1019 uint64_t tcps_sc_sendcookie; /* SYN cookie sent */
1020 uint64_t tcps_sc_recvcookie; /* SYN cookie received */
1021 uint64_t tcps_sc_spurcookie; /* SYN cookie spurious, rejected */
1022 uint64_t tcps_sc_failcookie; /* SYN cookie failed, rejected */
1023
1024 uint64_t tcps_hc_added; /* entry added to hostcache */
1025 uint64_t tcps_hc_bucketoverflow;/* hostcache per bucket limit hit */
1026
1027 uint64_t tcps_finwait2_drops; /* Drop FIN_WAIT_2 connection after time limit */
1028
1029 /* SACK related stats */
1030 uint64_t tcps_sack_recovery_episode; /* SACK recovery episodes */
1031 uint64_t tcps_sack_rexmits; /* SACK rexmit segments */
1032 uint64_t tcps_sack_rexmits_tso; /* SACK rexmit TSO chunks */
1033 uint64_t tcps_sack_rexmit_bytes; /* SACK rexmit bytes */
1034 uint64_t tcps_sack_rcv_blocks; /* SACK blocks (options) received */
1035 uint64_t tcps_sack_send_blocks; /* SACK blocks (options) sent */
1036 uint64_t tcps_sack_lostrexmt; /* SACK lost retransmission recovered */
1037 uint64_t tcps_sack_sboverflow; /* times scoreboard overflowed */
1038
1039 /* ECN related stats */
1040 uint64_t tcps_ecn_rcvce; /* ECN Congestion Experienced */
1041 uint64_t tcps_ecn_rcvect0; /* ECN Capable Transport */
1042 uint64_t tcps_ecn_rcvect1; /* ECN Capable Transport */
1043 uint64_t tcps_ecn_shs; /* ECN successful handshakes */
1044 uint64_t tcps_ecn_rcwnd; /* # times ECN reduced the cwnd */
1045
1046 /* TCP_SIGNATURE related stats */
1047 uint64_t tcps_sig_rcvgoodsig; /* Total matching signature received */
1048 uint64_t tcps_sig_rcvbadsig; /* Total bad signature received */
1049 uint64_t tcps_sig_err_buildsig; /* Failed to make signature */
1050 uint64_t tcps_sig_err_sigopt; /* No signature expected by socket */
1051 uint64_t tcps_sig_err_nosigopt; /* No signature provided by segment */
1052
1053 /* Path MTU Discovery Black Hole Detection related stats */
1054 uint64_t tcps_pmtud_blackhole_activated; /* Black Hole Count */
1055 uint64_t tcps_pmtud_blackhole_activated_min_mss; /* BH at min MSS Count */
1056 uint64_t tcps_pmtud_blackhole_failed; /* Black Hole Failure Count */
1057
1058 uint64_t tcps_tunneled_pkts; /* Packets encap's in UDP received */
1059 uint64_t tcps_tunneled_errs; /* Packets that had errors that were UDP encaped */
1060
1061 /* Dsack related stats */
1062 uint64_t tcps_dsack_count; /* Number of ACKs arriving with DSACKs */
1063 uint64_t tcps_dsack_bytes; /* Number of bytes DSACK'ed no TLP */
1064 uint64_t tcps_dsack_tlp_bytes; /* Number of bytes DSACK'ed due to TLPs */
1065
1066 /* TCPS_TIME_WAIT usage stats */
1067 uint64_t tcps_tw_recycles; /* Times time-wait was recycled. */
1068 uint64_t tcps_tw_resets; /* Times time-wait sent a reset. */
1069 uint64_t tcps_tw_responds; /* Times time-wait sent a valid ack. */
1070
1071 /* Accurate ECN Handshake stats */
1072 uint64_t tcps_ace_nect; /* ACE SYN packet with Non-ECT */
1073 uint64_t tcps_ace_ect1; /* ACE SYN packet with ECT1 */
1074 uint64_t tcps_ace_ect0; /* ACE SYN packet with ECT0 */
1075 uint64_t tcps_ace_ce; /* ACE SYN packet with CE */
1076
1077 /* ECN related stats */
1078 uint64_t tcps_ecn_sndect0; /* ECN Capable Transport */
1079 uint64_t tcps_ecn_sndect1; /* ECN Capable Transport */
1080
1081 /*
1082 * BBR and Rack implement TLP's these values count TLP bytes in
1083 * two catagories, bytes that were retransmitted and bytes that
1084 * were newly transmited. Both types can serve as TLP's but they
1085 * are accounted differently.
1086 */
1087 uint64_t tcps_tlpresends; /* number of tlp resends */
1088 uint64_t tcps_tlpresend_bytes; /* number of bytes resent by tlp */
1089
1090 /* SEG.ACK validation failures */
1091 uint64_t tcps_rcvghostack; /* received ACK for data never sent */
1092 uint64_t tcps_rcvacktooold; /* received ACK for data too long ago */
1093
1094
1095 uint64_t _pad[1]; /* 1 TBD placeholder for STABLE */
1096};
1097
1098#define tcps_rcvmemdrop tcps_rcvreassfull /* compat */
1099
1100#ifdef _KERNEL
1101#include <sys/counter.h>
1102#include <netinet/in_kdtrace.h>
1103
1104VNET_PCPUSTAT_DECLARE(struct tcpstat, tcpstat); /* tcp statistics */
1105/*
1106 * In-kernel consumers can use these accessor macros directly to update
1107 * stats.
1108 */
1109#define TCPSTAT_ADD(name, val) \
1110 do { \
1111 MIB_SDT_PROBE1(tcp, count, name, (val)); \
1112 VNET_PCPUSTAT_ADD(struct tcpstat, tcpstat, name, (val)); \
1113 } while (0)
1114#define TCPSTAT_INC(name) TCPSTAT_ADD(name, 1)
1115
1116/*
1117 * Kernel module consumers must use this accessor macro.
1118 */
1119void kmod_tcpstat_add(int statnum, int val);
1120#define KMOD_TCPSTAT_ADD(name, val) \
1121 do { \
1122 MIB_SDT_PROBE1(tcp, count, name, (val)); \
1123 kmod_tcpstat_add(offsetof(struct tcpstat, name) / \
1124 sizeof(uint64_t), \
1125 val); \
1126 } while (0)
1127#define KMOD_TCPSTAT_INC(name) KMOD_TCPSTAT_ADD(name, 1)
1128
1129/*
1130 * Running TCP connection count by state.
1131 */
1132VNET_DECLARE(counter_u64_t, tcps_states[TCP_NSTATES]);
1133#define V_tcps_states VNET(tcps_states)
1134#define TCPSTATES_INC(state) counter_u64_add(V_tcps_states[state], 1)
1135#define TCPSTATES_DEC(state) counter_u64_add(V_tcps_states[state], -1)
1136
1137/*
1138 * TCP specific helper hook point identifiers.
1139 */
1140#define HHOOK_TCP_EST_IN 0
1141#define HHOOK_TCP_EST_OUT 1
1142#define HHOOK_TCP_LAST HHOOK_TCP_EST_OUT
1143
1144struct tcp_hhook_data {
1145 struct tcpcb *tp;
1146 struct tcphdr *th;
1147 struct tcpopt *to;
1148 uint32_t len;
1149 int tso;
1150 tcp_seq curack;
1151};
1152#ifdef TCP_HHOOK
1153void hhook_run_tcp_est_out(struct tcpcb *tp,
1154 struct tcphdr *th, struct tcpopt *to,
1155 uint32_t len, int tso);
1156#endif
1157#endif
1158
1159/*
1160 * TCB structure exported to user-land via sysctl(3).
1161 *
1162 * Fields prefixed with "xt_" are unique to the export structure, and fields
1163 * with "t_" or other prefixes match corresponding fields of 'struct tcpcb'.
1164 *
1165 * Legend:
1166 * (s) - used by userland utilities in src
1167 * (p) - used by utilities in ports
1168 * (3) - is known to be used by third party software not in ports
1169 * (n) - no known usage
1170 *
1171 * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
1172 * included. Not all of our clients do.
1173 */
1174#if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_)
1175struct xtcpcb {
1176 ksize_t xt_len; /* length of this structure */
1177 struct xinpcb xt_inp;
1178 char xt_stack[TCP_FUNCTION_NAME_LEN_MAX]; /* (s) */
1179 char xt_logid[TCP_LOG_ID_LEN]; /* (s) */
1180 char xt_cc[TCP_CA_NAME_MAX]; /* (s) */
1181 int64_t spare64[6];
1182 int32_t t_state; /* (s,p) */
1183 uint32_t t_flags; /* (s,p) */
1184 int32_t t_sndzerowin; /* (s) */
1185 int32_t t_sndrexmitpack; /* (s) */
1186 int32_t t_rcvoopack; /* (s) */
1187 int32_t t_rcvtime; /* (s) */
1188 int32_t tt_rexmt; /* (s) */
1189 int32_t tt_persist; /* (s) */
1190 int32_t tt_keep; /* (s) */
1191 int32_t tt_2msl; /* (s) */
1192 int32_t tt_delack; /* (s) */
1193 int32_t t_logstate; /* (3) */
1194 uint32_t t_snd_cwnd; /* (s) */
1195 uint32_t t_snd_ssthresh; /* (s) */
1196 uint32_t t_maxseg; /* (s) */
1197 uint32_t t_rcv_wnd; /* (s) */
1198 uint32_t t_snd_wnd; /* (s) */
1199 uint32_t xt_ecn; /* (s) */
1200 uint32_t t_dsack_bytes; /* (n) */
1201 uint32_t t_dsack_tlp_bytes; /* (n) */
1202 uint32_t t_dsack_pack; /* (n) */
1203 uint16_t xt_encaps_port; /* (s) */
1204 int16_t spare16;
1205 int32_t spare32[22];
1206} __aligned(8);
1207
1208#ifdef _KERNEL
1209void tcp_inptoxtp(const struct inpcb *, struct xtcpcb *);
1210#endif
1211#endif
1212
1213/*
1214 * TCP function information (name-to-id mapping, aliases, and refcnt)
1215 * exported to user-land via sysctl(3).
1216 */
1217struct tcp_function_info {
1218 uint32_t tfi_refcnt;
1219 uint8_t tfi_id;
1220 char tfi_name[TCP_FUNCTION_NAME_LEN_MAX];
1221 char tfi_alias[TCP_FUNCTION_NAME_LEN_MAX];
1222};
1223
1224/*
1225 * Identifiers for TCP sysctl nodes
1226 */
1227#define TCPCTL_DO_RFC1323 1 /* use RFC-1323 extensions */
1228#define TCPCTL_MSSDFLT 3 /* MSS default */
1229#define TCPCTL_STATS 4 /* statistics */
1230#define TCPCTL_RTTDFLT 5 /* default RTT estimate */
1231#define TCPCTL_KEEPIDLE 6 /* keepalive idle timer */
1232#define TCPCTL_KEEPINTVL 7 /* interval to send keepalives */
1233#define TCPCTL_SENDSPACE 8 /* send buffer space */
1234#define TCPCTL_RECVSPACE 9 /* receive buffer space */
1235#define TCPCTL_KEEPINIT 10 /* timeout for establishing syn */
1236#define TCPCTL_PCBLIST 11 /* list of all outstanding PCBs */
1237#define TCPCTL_DELACKTIME 12 /* time before sending delayed ACK */
1238#define TCPCTL_V6MSSDFLT 13 /* MSS default for IPv6 */
1239#define TCPCTL_SACK 14 /* Selective Acknowledgement,rfc 2018 */
1240#define TCPCTL_DROP 15 /* drop tcp connection */
1241#define TCPCTL_STATES 16 /* connection counts by TCP state */
1242#define TCPCTL_KTLSLIST 17 /* connections with active ktls
1243 session */
1244#define TCPCTL_KTLSLIST_WKEYS 18 /* KTLSLIST with key data exported */
1245
1246#ifdef _KERNEL
1247#ifdef SYSCTL_DECL
1248SYSCTL_DECL(_net_inet_tcp);
1249SYSCTL_DECL(_net_inet_tcp_sack);
1250MALLOC_DECLARE(M_TCPLOG);
1251#endif
1252
1253VNET_DECLARE(int, tcp_log_in_vain);
1254#define V_tcp_log_in_vain VNET(tcp_log_in_vain)
1255
1256/*
1257 * Global TCP tunables shared between different stacks.
1258 * Please keep the list sorted.
1259 */
1260VNET_DECLARE(int, drop_synfin);
1261VNET_DECLARE(int, path_mtu_discovery);
1262VNET_DECLARE(int, tcp_abc_l_var);
1263VNET_DECLARE(uint32_t, tcp_ack_war_cnt);
1264VNET_DECLARE(uint32_t, tcp_ack_war_time_window);
1265VNET_DECLARE(int, tcp_autorcvbuf_max);
1266VNET_DECLARE(int, tcp_autosndbuf_inc);
1267VNET_DECLARE(int, tcp_autosndbuf_max);
1268VNET_DECLARE(int, tcp_bind_all_fibs);
1269VNET_DECLARE(int, tcp_delack_enabled);
1270VNET_DECLARE(int, tcp_do_autorcvbuf);
1271VNET_DECLARE(int, tcp_do_autosndbuf);
1272VNET_DECLARE(int, tcp_do_ecn);
1273VNET_DECLARE(int, tcp_do_lrd);
1274VNET_DECLARE(int, tcp_do_prr);
1275VNET_DECLARE(int, tcp_do_prr_conservative);
1276VNET_DECLARE(int, tcp_do_newcwv);
1277VNET_DECLARE(int, tcp_do_rfc1323);
1278VNET_DECLARE(int, tcp_tolerate_missing_ts);
1279VNET_DECLARE(int, tcp_do_rfc3042);
1280VNET_DECLARE(int, tcp_do_rfc3390);
1281VNET_DECLARE(int, tcp_do_rfc3465);
1282VNET_DECLARE(int, tcp_do_newsack);
1283VNET_DECLARE(int, tcp_do_sack);
1284VNET_DECLARE(int, tcp_do_tso);
1285VNET_DECLARE(int, tcp_ecn_maxretries);
1286VNET_DECLARE(int, tcp_initcwnd_segments);
1287VNET_DECLARE(int, tcp_insecure_rst);
1288VNET_DECLARE(int, tcp_insecure_syn);
1289VNET_DECLARE(int, tcp_insecure_ack);
1290VNET_DECLARE(uint32_t, tcp_map_entries_limit);
1291VNET_DECLARE(uint32_t, tcp_map_split_limit);
1292VNET_DECLARE(int, tcp_minmss);
1293VNET_DECLARE(int, tcp_mssdflt);
1294#ifdef STATS
1295VNET_DECLARE(int, tcp_perconn_stats_dflt_tpl);
1296VNET_DECLARE(int, tcp_perconn_stats_enable);
1297#endif /* STATS */
1298VNET_DECLARE(int, tcp_recvspace);
1299VNET_DECLARE(int, tcp_retries);
1300VNET_DECLARE(int, tcp_sack_globalholes);
1301VNET_DECLARE(int, tcp_sack_globalmaxholes);
1302VNET_DECLARE(int, tcp_sack_maxholes);
1303VNET_DECLARE(int, tcp_sack_tso);
1304VNET_DECLARE(int, tcp_sc_rst_sock_fail);
1305VNET_DECLARE(int, tcp_sendspace);
1306VNET_DECLARE(int, tcp_udp_tunneling_overhead);
1307VNET_DECLARE(int, tcp_udp_tunneling_port);
1308VNET_DECLARE(struct inpcbinfo, tcbinfo);
1309
1310#define V_tcp_do_lrd VNET(tcp_do_lrd)
1311#define V_tcp_do_prr VNET(tcp_do_prr)
1312#define V_tcp_do_newcwv VNET(tcp_do_newcwv)
1313#define V_drop_synfin VNET(drop_synfin)
1314#define V_path_mtu_discovery VNET(path_mtu_discovery)
1315#define V_tcbinfo VNET(tcbinfo)
1316#define V_tcp_abc_l_var VNET(tcp_abc_l_var)
1317#define V_tcp_ack_war_cnt VNET(tcp_ack_war_cnt)
1318#define V_tcp_ack_war_time_window VNET(tcp_ack_war_time_window)
1319#define V_tcp_autorcvbuf_max VNET(tcp_autorcvbuf_max)
1320#define V_tcp_autosndbuf_inc VNET(tcp_autosndbuf_inc)
1321#define V_tcp_autosndbuf_max VNET(tcp_autosndbuf_max)
1322#define V_tcp_bind_all_fibs VNET(tcp_bind_all_fibs)
1323#define V_tcp_delack_enabled VNET(tcp_delack_enabled)
1324#define V_tcp_do_autorcvbuf VNET(tcp_do_autorcvbuf)
1325#define V_tcp_do_autosndbuf VNET(tcp_do_autosndbuf)
1326#define V_tcp_do_ecn VNET(tcp_do_ecn)
1327#define V_tcp_do_rfc1323 VNET(tcp_do_rfc1323)
1328#define V_tcp_tolerate_missing_ts VNET(tcp_tolerate_missing_ts)
1329#define V_tcp_ts_offset_per_conn VNET(tcp_ts_offset_per_conn)
1330#define V_tcp_do_rfc3042 VNET(tcp_do_rfc3042)
1331#define V_tcp_do_rfc3390 VNET(tcp_do_rfc3390)
1332#define V_tcp_do_rfc3465 VNET(tcp_do_rfc3465)
1333#define V_tcp_do_newsack VNET(tcp_do_newsack)
1334#define V_tcp_do_sack VNET(tcp_do_sack)
1335#define V_tcp_do_tso VNET(tcp_do_tso)
1336#define V_tcp_ecn_maxretries VNET(tcp_ecn_maxretries)
1337#define V_tcp_initcwnd_segments VNET(tcp_initcwnd_segments)
1338#define V_tcp_insecure_rst VNET(tcp_insecure_rst)
1339#define V_tcp_insecure_syn VNET(tcp_insecure_syn)
1340#define V_tcp_insecure_ack VNET(tcp_insecure_ack)
1341#define V_tcp_map_entries_limit VNET(tcp_map_entries_limit)
1342#define V_tcp_map_split_limit VNET(tcp_map_split_limit)
1343#define V_tcp_minmss VNET(tcp_minmss)
1344#define V_tcp_mssdflt VNET(tcp_mssdflt)
1345#ifdef STATS
1346#define V_tcp_perconn_stats_dflt_tpl VNET(tcp_perconn_stats_dflt_tpl)
1347#define V_tcp_perconn_stats_enable VNET(tcp_perconn_stats_enable)
1348#endif /* STATS */
1349#define V_tcp_recvspace VNET(tcp_recvspace)
1350#define V_tcp_retries VNET(tcp_retries)
1351#define V_tcp_sack_globalholes VNET(tcp_sack_globalholes)
1352#define V_tcp_sack_globalmaxholes VNET(tcp_sack_globalmaxholes)
1353#define V_tcp_sack_maxholes VNET(tcp_sack_maxholes)
1354#define V_tcp_sack_tso VNET(tcp_sack_tso)
1355#define V_tcp_sc_rst_sock_fail VNET(tcp_sc_rst_sock_fail)
1356#define V_tcp_sendspace VNET(tcp_sendspace)
1357#define V_tcp_udp_tunneling_overhead VNET(tcp_udp_tunneling_overhead)
1358#define V_tcp_udp_tunneling_port VNET(tcp_udp_tunneling_port)
1359
1360#ifdef TCP_HHOOK
1361VNET_DECLARE(struct hhook_head *, tcp_hhh[HHOOK_TCP_LAST + 1]);
1362#define V_tcp_hhh VNET(tcp_hhh)
1363#endif
1364
1365void tcp_account_for_send(struct tcpcb *, uint32_t, uint8_t, uint8_t, bool);
1366int tcp_addoptions(struct tcpopt *, u_char *);
1367struct tcpcb *
1368 tcp_close(struct tcpcb *);
1369void tcp_discardcb(struct tcpcb *);
1370void tcp_twstart(struct tcpcb *);
1371int tcp_ctloutput(struct socket *, struct sockopt *);
1372void tcp_fini(void *);
1373char *tcp_log_addrs(struct in_conninfo *, struct tcphdr *, const void *,
1374 const void *);
1375char *tcp_log_vain(struct in_conninfo *, struct tcphdr *, const void *,
1376 const void *);
1377int tcp_reass(struct tcpcb *, struct tcphdr *, tcp_seq *, int *,
1378 struct mbuf *);
1379void tcp_reass_global_init(void);
1380void tcp_reass_flush(struct tcpcb *);
1381void tcp_dooptions(struct tcpopt *, u_char *, int, int);
1382void tcp_dropwithreset(struct mbuf *, struct tcphdr *, struct tcpcb *, int);
1383void tcp_pulloutofband(struct socket *,
1384 struct tcphdr *, struct mbuf *, int);
1385void tcp_xmit_timer(struct tcpcb *, int);
1386void tcp_newreno_partial_ack(struct tcpcb *, struct tcphdr *);
1387void cc_ack_received(struct tcpcb *tp, struct tcphdr *th,
1388 uint16_t nsegs, uint16_t type);
1389void cc_conn_init(struct tcpcb *tp);
1390void cc_post_recovery(struct tcpcb *tp, struct tcphdr *th);
1391void cc_ecnpkt_handler(struct tcpcb *tp, struct tcphdr *th, uint8_t iptos);
1392void cc_ecnpkt_handler_flags(struct tcpcb *tp, uint16_t flags, uint8_t iptos);
1393void cc_cong_signal(struct tcpcb *tp, struct tcphdr *th, uint32_t type);
1394#ifdef TCP_HHOOK
1395void hhook_run_tcp_est_in(struct tcpcb *tp,
1396 struct tcphdr *th, struct tcpopt *to);
1397#endif
1398
1399int tcp_input(struct mbuf **, int *, int);
1400int tcp_autorcvbuf(struct mbuf *, struct tcphdr *, struct socket *,
1401 struct tcpcb *, int);
1402int tcp_input_with_port(struct mbuf **, int *, int, uint16_t);
1403void tcp_do_segment(struct tcpcb *, struct mbuf *, struct tcphdr *, int,
1404 int, uint8_t);
1405
1406int register_tcp_functions(struct tcp_function_block *blk, int wait);
1407int register_tcp_functions_as_names(struct tcp_function_block *blk,
1408 int wait, const char *names[], int *num_names);
1409int register_tcp_functions_as_name(struct tcp_function_block *blk,
1410 const char *name, int wait);
1411int deregister_tcp_functions(struct tcp_function_block *blk, bool quiesce,
1412 bool force);
1413struct tcp_function_block *find_and_ref_tcp_functions(struct tcp_function_set *fs);
1414int find_tcp_function_alias(struct tcp_function_block *blk, struct tcp_function_set *fs);
1415uint32_t tcp_get_srtt(struct tcpcb *tp, int granularity);
1416void tcp_switch_back_to_default(struct tcpcb *tp);
1417struct tcp_function_block *
1418find_and_ref_tcp_fb(struct tcp_function_block *fs);
1419int tcp_default_ctloutput(struct tcpcb *tp, struct sockopt *sopt);
1420int tcp_ctloutput_set(struct inpcb *inp, struct sockopt *sopt);
1421void tcp_log_socket_option(struct tcpcb *tp, uint32_t option_num,
1422 uint32_t option_val, int err);
1423
1424
1425extern counter_u64_t tcp_inp_lro_direct_queue;
1426extern counter_u64_t tcp_inp_lro_wokeup_queue;
1427extern counter_u64_t tcp_inp_lro_compressed;
1428extern counter_u64_t tcp_inp_lro_locks_taken;
1429extern counter_u64_t tcp_extra_mbuf;
1430extern counter_u64_t tcp_would_have_but;
1431extern counter_u64_t tcp_comp_total;
1432extern counter_u64_t tcp_uncomp_total;
1433extern counter_u64_t tcp_bad_csums;
1434
1435extern uint32_t tcp_ack_war_time_window;
1436extern uint32_t tcp_ack_war_cnt;
1437
1438/*
1439 * Used by tcp_maxmtu() to communicate interface specific features
1440 * and limits at the time of connection setup.
1441 */
1442struct tcp_ifcap {
1443 int ifcap;
1444 u_int tsomax;
1445 u_int tsomaxsegcount;
1446 u_int tsomaxsegsize;
1447 bool ipsec_tso;
1448};
1449uint32_t tcp_maxmtu(struct in_conninfo *, struct tcp_ifcap *);
1450uint32_t tcp_maxmtu6(struct in_conninfo *, struct tcp_ifcap *);
1451
1452void tcp6_use_min_mtu(struct tcpcb *);
1453u_int tcp_maxseg(const struct tcpcb *);
1454u_int tcp_fixed_maxseg(const struct tcpcb *);
1455void tcp_mss_update(struct tcpcb *, int, int, struct hc_metrics_lite *,
1456 struct tcp_ifcap *);
1457void tcp_mss(struct tcpcb *, int);
1458int tcp_mssopt(struct in_conninfo *);
1459struct tcpcb *
1460 tcp_newtcpcb(struct inpcb *, struct tcpcb *);
1461int tcp_default_output(struct tcpcb *);
1462void tcp_state_change(struct tcpcb *, int);
1463void tcp_respond(struct tcpcb *, void *,
1464 struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, uint16_t);
1465bool tcp_challenge_ack_check(sbintime_t *, uint32_t *);
1466void tcp_send_challenge_ack(struct tcpcb *, struct tcphdr *, struct mbuf *);
1467bool tcp_twcheck(struct inpcb *, struct tcpopt *, struct tcphdr *,
1468 struct mbuf *, int);
1469void tcp_setpersist(struct tcpcb *);
1470void tcp_record_dsack(struct tcpcb *tp, tcp_seq start, tcp_seq end, int tlp);
1471struct tcptemp *
1472 tcpip_maketemplate(struct inpcb *);
1473void tcpip_fillheaders(struct inpcb *, uint16_t, void *, void *);
1474void tcp_timer_activate(struct tcpcb *, tt_which, u_int);
1475bool tcp_timer_active(struct tcpcb *, tt_which);
1476void tcp_timer_stop(struct tcpcb *);
1477int inp_to_cpuid(struct inpcb *inp);
1478/*
1479 * All tcp_hc_* functions are IPv4 and IPv6 (via in_conninfo)
1480 */
1481void tcp_hc_init(void);
1482#ifdef VIMAGE
1483void tcp_hc_destroy(void);
1484#endif
1485void tcp_hc_get(const struct in_conninfo *, struct hc_metrics_lite *);
1486uint32_t tcp_hc_getmtu(const struct in_conninfo *);
1487void tcp_hc_updatemtu(const struct in_conninfo *, uint32_t);
1488void tcp_hc_update(const struct in_conninfo *, struct hc_metrics_lite *);
1489void cc_after_idle(struct tcpcb *tp);
1490
1491extern struct protosw tcp_protosw; /* shared for TOE */
1492extern struct protosw tcp6_protosw; /* shared for TOE */
1493
1494uint32_t tcp_new_ts_offset(struct in_conninfo *);
1495tcp_seq tcp_new_isn(struct in_conninfo *);
1496
1497sackstatus_t
1498 tcp_sack_doack(struct tcpcb *, struct tcpopt *, tcp_seq);
1499int tcp_dsack_block_exists(struct tcpcb *);
1500void tcp_update_dsack_list(struct tcpcb *, tcp_seq, tcp_seq);
1501void tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_laststart,
1502 tcp_seq rcv_lastend);
1503void tcp_clean_dsack_blocks(struct tcpcb *tp);
1504void tcp_clean_sackreport(struct tcpcb *tp);
1505int tcp_sack_adjust(struct tcpcb *tp);
1506struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt);
1507void tcp_do_prr_ack(struct tcpcb *, struct tcphdr *, struct tcpopt *,
1508 sackstatus_t, u_int *);
1509void tcp_lost_retransmission(struct tcpcb *, struct tcphdr *);
1510void tcp_sack_partialack(struct tcpcb *, struct tcphdr *, u_int *);
1511void tcp_resend_sackholes(struct tcpcb *tp);
1512void tcp_free_sackholes(struct tcpcb *tp);
1513void tcp_sack_lost_retransmission(struct tcpcb *, struct tcphdr *);
1514int tcp_newreno(struct tcpcb *, struct tcphdr *);
1515int tcp_compute_pipe(struct tcpcb *);
1516uint32_t tcp_compute_initwnd(uint32_t);
1517void tcp_sndbuf_autoscale(struct tcpcb *, struct socket *, uint32_t);
1518int tcp_stats_sample_rollthedice(struct tcpcb *tp, void *seed_bytes,
1519 size_t seed_len);
1520int tcp_can_enable_pacing(void);
1521int tcp_incr_dgp_pacing_cnt(void);
1522void tcp_dec_dgp_pacing_cnt(void);
1523void tcp_decrement_paced_conn(void);
1524void tcp_change_time_units(struct tcpcb *, int);
1525void tcp_handle_orphaned_packets(struct tcpcb *);
1526
1527struct mbuf *
1528 tcp_m_copym(struct mbuf *m, int32_t off0, int32_t *plen,
1529 int32_t seglimit, int32_t segsize, struct sockbuf *sb, bool hw_tls);
1530
1531int tcp_stats_init(void);
1532void tcp_log_end_status(struct tcpcb *tp, uint8_t status);
1533#ifdef TCP_REQUEST_TRK
1534void tcp_req_free_a_slot(struct tcpcb *tp, struct tcp_sendfile_track *ent);
1535struct tcp_sendfile_track *
1536tcp_req_find_a_req_that_is_completed_by(struct tcpcb *tp, tcp_seq th_ack, int *ip);
1537int tcp_req_check_for_comp(struct tcpcb *tp, tcp_seq ack_point);
1538int
1539tcp_req_is_entry_comp(struct tcpcb *tp, struct tcp_sendfile_track *ent, tcp_seq ack_point);
1540struct tcp_sendfile_track *
1541tcp_req_find_req_for_seq(struct tcpcb *tp, tcp_seq seq);
1542void
1543tcp_req_log_req_info(struct tcpcb *tp,
1544 struct tcp_sendfile_track *req, uint16_t slot,
1545 uint8_t val, uint64_t offset, uint64_t nbytes);
1546
1547uint32_t
1548tcp_estimate_tls_overhead(struct socket *so, uint64_t tls_usr_bytes);
1549void
1550tcp_req_alloc_req(struct tcpcb *tp, union tcp_log_userdata *user,
1551 uint64_t ts);
1552
1553struct tcp_sendfile_track *
1554tcp_req_alloc_req_full(struct tcpcb *tp, struct tcp_snd_req *req, uint64_t ts, int rec_dups);
1555
1556
1557#endif
1558#ifdef TCP_ACCOUNTING
1559int tcp_do_ack_accounting(struct tcpcb *tp, struct tcphdr *th, struct tcpopt *to, uint32_t tiwin, int mss);
1560#endif
1561
1562static inline void
1563tcp_lro_features_off(struct tcpcb *tp)
1564{
1565 tp->t_flags2 &= ~(TF2_SUPPORTS_MBUFQ|
1566 TF2_MBUF_QUEUE_READY|
1567 TF2_DONT_SACK_QUEUE|
1568 TF2_MBUF_ACKCMP|
1569 TF2_MBUF_L_ACKS);
1570}
1571
1572static inline void
1573tcp_fields_to_host(struct tcphdr *th)
1574{
1575
1576 th->th_seq = ntohl(th->th_seq);
1577 th->th_ack = ntohl(th->th_ack);
1578 th->th_win = ntohs(th->th_win);
1579 th->th_urp = ntohs(th->th_urp);
1580}
1581
1582static inline void
1583tcp_fields_to_net(struct tcphdr *th)
1584{
1585
1586 th->th_seq = htonl(th->th_seq);
1587 th->th_ack = htonl(th->th_ack);
1588 th->th_win = htons(th->th_win);
1589 th->th_urp = htons(th->th_urp);
1590}
1591#endif /* _KERNEL */
1592
1593#endif /* _NETINET_TCP_VAR_H_ */